Kolmogorov-Zurbenko filter coupled with machine learning to reveal multiple drivers of surface ozone pollution in China from 2015 to 2022

Abstract

Near-surface ozone pollution is a significant concern in China. Its meteorological drivers are uncontrolled, stressing an urgent need to quantify the anthropogenic-driven components. This study employs explainable machine learning (ML) algorithms to predict ozone concentrations, emphasizing the anthropogenic-driven trends after accounting for meteorological effects. Results present that radiation is the most important meteorological factor affecting ozone pollution in the Yangtze River Delta (YRD), Pearl River Delta (PRD), and Sichuan Basin (SCB), while temperature is dominant for the North China Plain (NCP). Even at lower temperatures, stronger solar radiation can still lead to the accumulation of higher ozone concentrations. The anthropogenic-driven ozone concentration showed an upward trend in China, with an interannual growth rate of 2.61 μg/m3 a−1 from 2015 to 2022. Nonetheless, its rising trend experienced a post-2019 downturn, due to the COVID-19 lockdown and emission reduction strategies. It started to rise in 2022. Regionally, NCP has the highest ozone concentration, the SCB has the most pronounced increase in 2022, but the PRD has no noticeable variation and no significant seasonal change after 2019. As for precursor emissions, the urban areas in China are mostly located in the VOC-limited (volatile organic component) and transitional regimes, highlighting that VOC control is more cost-effective.